Ralph Loop

This is a structured workflow for having AI agents build features autonomously until they meet your acceptance criteria. You write user stories as JSON files with testable conditions, then run agents in a loop where each one tries to satisfy the criteria and logs what happened for the next agent. It's an opinionated setup that includes agent configuration, story templates, and the actual loop runner called Ralph. The interesting part is treating AI development like CI/CD where passing tests equals done, though you'll need solid acceptance criteria upfront or the loop just optimizes for the wrong thing. Best for repetitive feature work where you can define clear success conditions.